2023 Volume 30 Issue 10 Pages 1492-1506
Aims: To examine the trajectory of white blood cell (WBC) and their potential impacts on cardiovascular disease (CVD) and all-cause mortality (ACM) risks.
Methods: This prospective cohort included 61,666 participants without CVD on or before June 1, 2012. Latent mixture modeling was used to identify WBC trajectories in 2006-2012 as predictors of CVD and ACM. Incident CVD and ACM in 2012-2019 were the outcomes. Cox proportional hazards models were fitted to analyze the risks of incident CVD and ACM.
Results: According to WBC ranges and dynamics, five distinct WBC trajectories were identified: low-stable (n=18,432), moderate-stable (n=26,656), elevated-stable (n=3,153), moderate-increasing (n=11,622), and elevated-decreasing (n=1,803). During 6.65±0.83 years of follow-up, we documented 3773 incident CVD cases and 3304 deaths. Relative to the low-stable pattern, the moderate-increasing pattern was predictive of an elevated risk of CVD (HR=1.36, 95% CI: 1.24-1.50), especially acute myocardial infarction (AMI) (HR=1.91, 95% CI: 1.46-2.51), while the elevated-stable pattern was predictive of an elevated risk of ACM (HR=1.77, 95% CI: 1.52-2.06). Among participants with hs-CRP ＜2 mg/L or ≥2 mg/L, similar associations were observed between the moderate-increasing pattern with CVD (HR=1.41, 95% CI: 1.24-1.61) and ACM (HR=1.54, 95% CI: 1.18-2.01, HR=1.89, 95% CI: 1.57-2.29, respectively).
Conclusions: We found that distinct WBC trajectories were differentially associated with CVD and ACM risks in Chinese adults.
Cardiovascular disease (CVD) is the world’s leading cause of mortality, accounting for 31% of all deaths1). Even when recognized CVD risk factors, including hypertension, diabetes, dyslipidemia, and other factors, are controlled, CVD risk is not eliminated, resulting in residual risk2). Inflammation plays an important role in residual risk3-7). As shown in clinical trials8, 9), statins can lower the high-sensitivity C-reactive protein (hs-CRP) level, and the degree of hs-CRP reduction is proportionate to the reduced cardiovascular risk. The CANTOS study10) and the LODOCO-II trial11) provided further evidence that anti-inflammatory therapy can reduce the residual risk of CVD.
White blood cells (WBCs) are common biomarkers of inflammation. Links between WBC counts and risks of CVD and all-cause mortality were first reported by Grimm et al.12) in 1987. Since then, several studies have shown associations between WBC count and CVD as well as all-cause mortality in the general population13, 14), women15), and elderly individuals16, 17), although the findings are not consistent. The ARIC study18), the CIRCS study19), and other studies found that a high WBC count was a risk factor for ischemic stroke (IS). Negative results were reported by the Japanese cohort study20) and The Malmö Diet and Cancer Study21). However, WBC counts may vary over short or long periods owing to the influence of the environment, lifestyle, and certain inflammatory illnesses22). A single measurement cannot accurately represent the impacts of WBC counts on CVD and all-cause mortality. To remedy this deficiency, we constructed a trajectory model to investigate the impact of long-term WBC variations on the risks of CVD and all-cause mortality in the general population.
The Kailuan Study was a prospective dynamic cohort study conducted in Tangshan, China, to investigate risk factors for non-communicable chronic diseases, such as CVD. Please see The Kailuan Study, trial identification: ChiCTR-TNC-11001489, registration site: http://www.chictr.org.cn/index.aspx, registration number: 11001489 23), for further details. Since June 2006, 101,510 people over the age of 18 years have been included in the study and have completed questionnaires, clinical exams, and laboratory testing. Every two years, all participants are followed, and the incidence of chronic illnesses (such as CVD) and mortality is documented.
In the current study, WBC trajectories were developed from 2006 to 2012 to predict CVD and ACM risk from 2012 to 2019. In other words, the study was restricted to the population who participated in the examinations in 2006, 2008 and/or 2010, 2012 and had their WBC measurements taken biennially. Participants were excluded if they: (1) failed to take 2008 and 2010 examinations, (2) had missing information of WBC during 2006-2012, (3) lacked measurements of relevant confounders including age, sex, total cholesterol (TC, in mmol/L), triglyceride (TG, in mmol/L), body mass index (BMI, in kg/m2), fasting blood glucose (FBG, in mmol/L), dietary salt intake, marital status, sedentary lifestyle, educational background, tobacco consumption, alcohol drinking, physical exercise, diabetes mellitus and hypertension, and (4) had a history of CVD, leukemia, lymphoma, myelodysplastic syndrome, cirrhosis, malignant solid tumors and rheumatic diseases at baseline or were diagnosed with CVD, leukemia, lymphoma, myelodysplastic syndrome, cirrhosis, malignant solid tumors and rheumatic diseases during 2006 to 2012 (trajectory patterns). A total of 61,666 individuals were left in the final analyses and scheduled a follow-up (Fig.1). (The flow chart is shown in Supplemental Fig.1). The characteristics of the participants who were included and excluded from the present analysis are presented in Supplemental Table 1.
WBC=White Blood Cell.
The flow chart of study
| Characteristic | Included (n = 61,666) | Excluded (n = 39,844) | P |
|---|---|---|---|
| Age (y) | 55.33±11.81 | 56.97±12.68 | ＜0.0001 |
| Male (%) | 47,203 (76.55%) | 30,002 (84.77%) | ＜0.0001 |
| Hypertension (%) | 24,194 (39.23%) | 19,428 (54.89%) | ＜0.0001 |
| Diabetes (%) | 5,979 (9.7%) | 4,365 (12.33%) | ＜0.0001 |
| Current alcohol consumption (%) | 15,385 (24.95%) | 11,357 (32.09%) | ＜0.0001 |
| Current smoking (%) | 15,155 (24.58%) | 11,605 (32.79%) | ＜0.0001 |
| Higher education level (%) | 13,327 (21.61%) | 5,174 (14.62%) | ＜0.0001 |
| Physical activity ≥ 3 times/week (%) | 9,342 (15.15%) | 6,029 (17.03%) | ＜0.0001 |
| Use of antihypertensive agents (%) | 5,151 (8.35%) | 5,371 (15.18%) | ＜0.0001 |
| Use of hypoglycemic agents (%) | 1,108 (1.80%) | 1,279 (3.61%) | ＜0.0001 |
| Use of lipid-lowering agents (%) | 397 (0.64%) | 457 (1.29%) | ＜0.0001 |
| SBP (mmHg) | 131.27±19.07 | 135.7±22.45 | ＜0.0001 |
| DBP (mmHg) | 83.88±10.46 | 84.88±12.28 | ＜0.0001 |
| FBG (mmol/L) | 5.75±2.07 | 5.62±1.95 | ＜0.0001 |
| TC (mmol/L) | 5.11±1.58 | 4.96±1.17 | ＜0.0001 |
| BMI (kg/m2) | 25.03±3.3 | 25.03±3.57 | ＜0.0001 |
| Hs-CRP [mg/L, M (P25, P75)] | 1.6 (0.80, 3.00) | 2.82 (1.87, 3.98) | ＜0.0001 |
| WBC count at 2006 (109/L) | 6.55±1.57 | 7.37±1.67 | ＜0.0001 |
Abbreviation: BMI, body mass index; DBP, diastolic blood pressure; SBP, systolic blood pressure; TC, total cholesterol; FPG, fasting plasma glucose.
Continuous variables are described as Means±SDs if normally distributed or medians (25th percentage, 75th percentage) if nonnormally distributed. Categorical variables are described as numbers (%). P values were derived from ANOVA tests for continuous variables and the chi- square test for categorical variables.
The study protocol was approved by the ethics committee of Kailuan General Hospital (NO.200605). All participants provided written informed consent.
Assessment of WBC Counts and TrajectoriesMorning fasting venous blood samples were collected from subjects to measure WBC counts, and all tests were conducted according to standard operating procedures. We used latent mixture modeling within the Proc Traj procedure24), which can identify distinctive groups of developmental trajectories within a population, to calculate changes in WBC counts from 2006 to 2012 as the exposure. The censored normal model in the PROC TRAJ procedure was applied. First, we grouped the model with five trajectory patterns, and then we chose the best-fitting model by comparing the Bayesian information criterion (BIC) of the five trajectory models to those with 4, 3, 2 and 1 trajectories. Second, we compared models with different functional forms. Cubic, quadratic and linear terms were considered and evaluated based on their significance level (p＜0.05), starting with the highest polynomial. In our final model, we had five trajectories with cubic order terms.
Assessment of Other CovariatesInformation about age, sex, smoking, alcohol consumption, physical activity, education level, and self-reported medical history (e.g., hypertension, diabetes, and active treatment such as hypoglycemic agents, antihypertensive agents and lipid-lowering agents) at baseline was collected via a questionnaire. Hypertension was defined as a self-reported history of hypertension, current treatment with an antihypertensive agent or a measured systolic blood pressure (SBP) ≥140 mmHg and/or diastolic blood pressure (DBP) ≥90 mmHg. Diabetes was defined as a self-reported history of diabetes, current treatment with a hypoglycemic agent or a fasting blood glucose (FBG) concentration ≥7.0 mmol/L. Hypercholesterolemia was defined as serum total cholesterol ≥5.2 mmol/L. Physical exercise was defined as an exercise frequency ≥3 times/week and duration ≥30 minutes/time. Alcohol consumption was defined as more than 0.1 serving/day, and smoking was defined as smoking at least one cigarette a day on average in the past year. Height, weight, and blood pressure were measured by trained field workers (i.e., nurses) during the surveys. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Total cholesterol (TC), FBG, and hs-CRP were assessed by an autoanalyzer (Hitachi 747, Hitachi, Tokyo, Japan) at the central laboratory of Kailuan Hospital. ΔWBC is the difference between the WBC count in 2012 and 2006 and was used to represent the change in WBC count from 2006 to 2012.
Ascertainment of Incident CVD Events and All-Cause MortalityThe participants were followed after completing the last examinations in 2012-2013 until the occurrence of CVD, death, or June 1, 2020, whichever occurred first. CVD, including acute myocardial infarction (AMI), heart failure (HF), and stroke, was diagnosed according to previously described criteria25). In brief, all participants in the Kailuan Study were included in the Social Insurance Institution and the Hospital Discharge Register, which was consulted to determine the incidence of CVD and all-cause mortality. To further identify potential CVD events, we reviewed the discharge lists from the 11 hospitals from 2006 to 2020 and asked for a history of CVD via a questionnaire during the biennial interview. For all suspected CVD events, 3 experienced physician adjudicators who were blinded to the study design reviewed the medical records. Incident AMI and HF were diagnosed according to the World Health Organization’s Multinational Monitoring of Trends and Determinants in Cardiovascular Disease criteria26). Stroke was diagnosed according to the World Health Organization’s criteria27), as reported previously28). Mortality information was collected from provincial vital statistics offices. Study clinicians (S.W. and Y.W.) reviewed all death certificates.
Statistical AnalysisThe general characteristics of the subjects are expressed as means and standard deviations (SDs) for continuous variables and as percentages for categorical variables. The hazard ratios (HRs) and 95% confidence intervals (CIs) for incident CVD and all-cause mortality according to the five WBC trajectories were analyzed by Cox proportional hazards regression models. The low-stable group was chosen as the reference group because previous studies13, 14, 18, 19) indicated that those with a low but normal WBC count had the lowest risk of CVD events and all-cause mortality. We assessed the proportional assumptions of the Cox proportional hazards regression models by (i) Schoenfeld residuals, (ii) the inclusion of a time-varying covariate(the interaction between the five trajectory groups and the logarithm of follow-up time) in PROC PHREG and (iii) the ASSESS statement. According to these three tests, our models met the proportional assumption criteria.
A Cox proportional hazards model was used to investigate the associations between exposures (WBC trajectories) and the risks of CVD development and death. The results were adjusted for age (continuous), sex (male or female), smoking (ever or never), alcohol consumption (ever or never),exercise level (＜3 or ≥3 times/week), educational level (senior high school or above or less than senior high school), hypertension (yes or no), diabetes(yes or no), hypercholesterolemia (yes or no), BMI (＜24 or ≥24 kg/m2), hypoglycemic agent use (yes or no), antihypertensive agent use (yes or no), lipid-lowering agent use (yes or no) and hs-CRP concentration(mean value of 4 measurements during the exposure period). Missing data for these covariates were coded as an extra category.
To examine whether the potential associations of WBC trajectories with CVD and all-cause mortality risk were due to the single measurement of WBCs, we further adjusted for WBC counts in 2006 and 2012 respectively, one at a time, in the Cox model. Because people with a higher concentration of hs-CRP (≥2 mg/L) are considered to have higher risks for CVD29) and all-cause mortality, likelihood ratio tests were conducted to examine statistical interactions between WBC trajectories and hs-CRP concentrations (＜2 or ≥2 mg/L)by comparing −2 log likelihood ×2 between nested models with and without the cross-product terms.
Sensitivity analysis of WBC trajectories and CVD and all-cause mortality risks were conducted as follows: (i) participants who developed CVD or died from any cause within 2 years after the measurement in 2012 were excluded to eliminate reverse causation and (ii) participants (21,531) missing one measurement during the exposure period were excluded. The multivariable-adjusted Fine-Gray regression model, considering death as a competing event, was used to calculate the risks of CVD and AMI not influenced by death.
For more detailed analyses of the association between changes in WBC counts from 2006 to 2012 and CVD and all-cause mortality risks, we also used restricted cubic spline (RCS) models with 3 knots defined at the 5th, 50th, and 95th percentiles of the ΔWBC.
The person-time of follow-up for each participant was from the date when the response was collected in the 2019 survey to either the date of CVD onset, death, loss to follow-up, or the end of follow-up (December 31, 2019). No participants were lost to follow-up.
All analyses were conducted using SAS version 9.3 (SAS Institute, Inc., Cary, NC). A two-sided p＜0.05 was considered statistically significant in the current study.
We categorized the study population into five observed discrete WBC trajectory groups based on WBC counts and changing patterns in the exposure period (Fig.2): 29.89% (n=18,432) of participants maintained a low WBC count (mean WBC counts ranged from 5.14×109/L to 5.09×109/L in 2006–2012, referred to as the “low-stable pattern”), 43.23% (n=26,656) of participants consistently maintained a moderate WBC count (mean WBC counts ranged from 6.49×109/L to 6.40×109/L in 2006–2012, referred to as the “moderate-stable pattern”), 5.11% (n=3,153) of participants consistently maintained an elevated WBC count (mean WBC counts ranged from 9.59×109/L to 9.1×109/L in 2006–2012, referred to as the “elevated-stable pattern”), 18.85% (n=11,622) of participants initially had a medium WBC counts that increased over time (mean WBC counts from 7.6×109/L in 2006 to 7.86×109/L in 2012, referred to as the “moderate-increasing pattern”), 2.92% (n=1,803) of participants initially had an elevated WBC count that later decreased over time (mean WBC counts decreased from 9.9×109/L in 2006 to 6.3×109/L in 2012, referred to as the “elevated-decreasing pattern”).
Dynamic trajectories of WBC counts during the exposure period (2006-2013) in the study population.
Among the 61,666 participants included in the current study, the mean age was 55.33±11.81 years, and the prevalence rates of hypertension and diabetes were 60.93% and 9.7%, respectively. Demographics data of the five trajectory groups are provided in Table 1. Age, sex, level of education, medical history, i.e., hypertension and diabetes, and biological data, including BMI, SBP, DBP, FBG, and TC, were notably different among the trajectory groups (Table 1).
| Characteristic | Low- stable | Moderate- stable | Elevated- stable | Moderate- increasing | Elevated- decreasing | P |
|---|---|---|---|---|---|---|
| N (%) | 18432 (29.89%) | 26656 (43.23%) | 3153 (5.11%) | 11622 (18.85%) | 1803 (2.92%) | |
| Age (y) | 55.84±11.75 | 55.39±11.9 | 53.92±11.33 | 55.03±11.81 | 53.4±11.51 | ＜0.0001 |
| Male (%) | 12426 (67.42%) | 20633 (77.4%) | 2805 (88.96%) | 9853 (84.78%) | 1486 (82.42%) | ＜0.0001 |
| Hypertension (%) | 8693 (47.16%) | 14596 (54.76%) | 1983 (62.89%) | 6946 (59.77%) | 983 (54.52%) | ＜0.0001 |
| Diabetes (%) | 1335 (7.24%) | 2602 (9.76%) | 477 (15.13%) | 1376 (11.84%) | 189 (10.48%) | ＜0.0001 |
| Current alcohol consumption (%) | 4199 (22.78%) | 6679 (25.06%) | 888 (28.16%) | 3141 (27.03%) | 478 (26.51%) | ＜0.0001 |
| Current smoking (%) | 3527 (19.14%) | 6478 (24.3%) | 1149 (36.44%) | 3493 (30.06%) | 508 (28.18%) | ＜0.0001 |
| Higher education level (%) | 4146 (22.49%) | 5710 (21.42%) | 696 (22.07%) | 2384 (20.51%) | 391 (21.69%) | ＜0.0001 |
| Physical activity ≥ 3 times/week (%) | 3068 (16.64%) | 4007 (15.03%) | 420 (13.32%) | 1634 (14.06%) | 213 (11.81%) | ＜0.0001 |
| Use of antihypertensive agents (%) | 2945 (15.98%) | 5275 (19.79%) | 764 (24.23%) | 2600 (22.37%) | 331 (18.36%) | ＜0.0001 |
| Use of hypoglycemic agents (%) | 795 (4.31%) | 1484 (5.57%) | 254 (8.06%) | 765 (6.58%) | 110 (6.1%) | ＜0.0001 |
| Use of lipid-lowering agents (%) | 1079 (5.85%) | 1981 (7.43%) | 272 (8.63%) | 942 (8.11%) | 165 (9.15%) | ＜0.0001 |
| SBP (mmHg) | 128.86±18.87 | 131.73±19.14 | 134.08±18.75 | 133.36±18.92 | 130.94±18.92 | ＜0.0001 |
| DBP (mmHg) | 82.45±10.3 | 84.18±10.51 | 85.5±10.26 | 85.06±10.41 | 83.69±10.49 | ＜0.0001 |
| FBG (mmol/L) | 5.6±1.88 | 5.75±1.97 | 6.04±2.49 | 5.9±2.43 | 5.75±1.8 | ＜0.0001 |
| TC (mmol/L) | 5.09±1.68 | 5.12±1.66 | 5.14±1.28 | 5.12±1.37 | 5.05±1.03 | ＜0.0001 |
| BMI (kg/m2) | 24.46±3.17 | 25.11±3.26 | 25.83±3.44 | 25.53±3.41 | 25.17±3.38 | ＜0.0001 |
| Hs-CRP [mg/L, M (P25, P75)]† | 1.37 (0.71, 2.58) | 1.62 (0.81, 2.99) | 2.05 (0.97, 3.78) | 1.83 (0.91, 3.39) | 1.83 (0.86, 3.47) | ＜0.0001 |
| WBC counts in 2006 (109/L) | 5.14±0.82 | 6.49±0.94 | 9.59±1.13 | 7.6±0.98 | 9.9±0.81 | ＜0.0001 |
| WBC counts in 2012 (109/L) | 5.08±0.78 | 6.4±0.95 | 9.1±1.14 | 7.86±1.08 | 6.63±1.11 | ＜0.0001 |
| ΔWBC [109/L, (P25, P75)] | －0.1 (－0.7, 0.6) | －0.1 (－1.0, 0.8) | －0.4 (－1.6, 0.7) | 0.2 (－0.8, 1.2) | －3.2 (－4.1, －2.4) | ＜0.0001 |
Abbreviation: BMI, body mass index, DBP, diastolic blood pressure, SBP, systolic blood pressure, TC, total cholesterol, FBG, fasting blood glucose. Continuous variables are described as Means±SDs if normally distributed or medians (25th percentage, 75th percentages) if nonnormally distributed. Categorical variables are described as numbers (%). P values were derived from ANOVA tests for continuous variables and chi-square test for category variables.
†Average concentrations based on measurements in 2006, 2008, 2010 and 2012.
After a median follow-up of 6.65 years, we observed 3773 total CVD events. The incidence rates in the low-stable pattern, moderate-stable pattern, elevated-stable pattern, moderate-increasing pattern, and elevated-decreasing pattern groups were 74.66, 93.12, 111.42, 116.65 and 87.09 per 105 person-years (PYs), respectively. Relative to those in the low-stable pattern group, participants in the moderate-increasing pattern group had a significantly higher cumulative rate of incident CVD events (Fig.3, log-rank test, p＜0.001).
(A) Participants with the moderate-increasing pattern had a significantly higher cumulative rate of CVD than those with the low-stable pattern (p＜0.0001).
(B) Participants with the elevated-stable pattern had significantly higher cumulative all-cause mortality than those with the low-stable pattern (p＜0.0001).
The association between the WBC trajectory and CVD risk is shown in Table 2. The moderate-increasing pattern was associated with the highest risk of the development of CVD. Relative to those for the low-stable pattern, the adjusted HRs were 1.36 (95% CI: 1.24-1.50) for the moderate-increasing pattern after adjustment for potential confounders.
| WBC trajectory patterns | |||||
|---|---|---|---|---|---|
| Low- stable | Moderate- stable | Elevated- stable | Moderate- increasing | Elevated- decreasing | |
| No. at risk (%) | 18,432 (29.89%) | 26,656 (43.23%) | 3,153 (5.11%) | 11,622 (18.85%) | 1,803 (2.92%) |
| CVD | |||||
| Cases, n (Incidence rate, per 105 person-years) | 911 (74.66) | 1637 (93.12) | 231 (111.42) | 890 (116.65) | 104 (87.09) |
| Multivariate Model 1a | 1 [Reference] | 1.16 (1.07-1.26) | 1.29 (1.12-1.50) | 1.36 (1.24-1.50) | 1.16 (0.94-1.42) |
| Multivariate Model 2b | 1 [Reference] | 1.12 (1.02-1.23) | 1.17 (0.95-1.45) | 1.29 (1.14-1.46) | 1.04 (0.80-1.35) |
| Multivariate Model 3c | 1 [Reference] | 1.09 (0.98-1.20) | 1.05 (0.84-1.31) | 1.21 (1.05-1.40) | 0.99 (0.78-1.25) |
| Sensitivity analysis-1 | 1 [Reference] | 1.10 (1.01-1.21) | 1.25 (1.05-1.47) | 1.30 (1.17-1.45) | 1.09 (0.87-1.38) |
| Sensitivity analysis-2 | 1 [Reference] | 1.26 (0.97-1.62) | 1.76 (1.49-2.09) | 1.31 (1.17-1.47) | 1.17 (1.06-1.29) |
| Fine-Gray model | 1 [Reference] | 1.45 (1.13-1.85) | 1.80 (1.21-2.67) | 1.91 (1.46-2.51) | 1.11 (0.59-2.07) |
| All-cause mortality | |||||
| Cases, n (incidence rate, per 105 person-years) | 821 (66.83) | 1443 (81.31) | 218 (104.05) | 742 (96.03) | 80 (66.23) |
| Multivariate Model 1a | 1 [Reference] | 1.18 (1.09-1.29) | 1.77 (1.52-2.06) | 1.42 (1.28-1.57) | 1.22 (0.97-1.54) |
| Multivariate Model 2b | 1 [Reference] | 1.16 (1.05-1.28) | 1.56 (1.25-1.96) | 1.35 (1.18-1.55) | 1.05 (0.78-1.41) |
| Multivariate Model 3c | 1 [Reference] | 1.09 (0.97-1.22) | 1.35 (1.06-1.73) | 1.22 (1.03-1.44) | 1.22 (0.94-1.59) |
| Sensitivity analysis-1 | 1 [Reference] | 1.17 (1.06-1.29) | 1.76 (1.49-2.09) | 1.31 (1.17-1.47) | 1.26 (0.97-1.62) |
| Sensitivity analysis-2 | 1 [Reference] | 1.27 (0.89-1.80) | 1.56 (1.23-1.98) | 1.43 (1.24-1.64) | 1.18 (1.05-1.33) |
Abbreviations: CVD, cardiovascular disease, AMI, acute myocardial infarction, IS, ischemic stroke, HS, hemorrhagic stroke, HR, hazard ratio.
a Model 1 adjusted for age, sex (men, women), smoking (ever or never smoker), alcohol consumption (ever or never drinker), exercise (＜3 or ≥ 3 times/week), educational level (senior high school and above or less than senior high school), hypertension (yes or no), diabetes (yes or no), hypercholesterolemia (yes or no), BMI in 2012 (＜24 or ≥ 24 kg/m2), hypoglycemic agent use (yes or no), antihypertensive agent use (yes or no), lipid-lowering agent use (yes or no) and mean serum concentration of high-sensitivity C-reactive protein during 2006-2010 (continuous).
b Model 2 adjusted for the factors in multivariate model 1 plus WBC in 2006, c Model 3 adjusted for the factors in multivariate model 1 plus WBC in 2012.
Sensitivity analysis-1: model 3 for participants after excluding 1,316 participants who developed CVD or died during 2012-2013,
Sensitivity analysis-2: model 3 for participants after excluding 21,531 participants with missing WBCs in 2008 or 2010.
After additional adjustment for the 2006 or 2012 WBC count, the association between the moderate-increasing pattern and the risk of CVD development remained robust, with HRs of 1.29 (95% CI, 1.14-1.46) and 1.21 (95% CI, 1.05-1.40), respectively. In the sensitivity analysis, the HRs for CVD associated with the moderate-increasing pattern remained robust after excluding participants with CVD that occurred during the first 2 years of follow-up and participants missing WBC data in 2008 or 2010.
RCS plots demonstrated significant linear associations of ΔWBC with CVD and AMI but no significant association of ΔWBC with stroke, IS or HS (Fig.4).
The associations of ΔWBC with incident CVD (a), AMI (b), all-cause mortality (c), Stroke (d), IS (e), and HS (f) were quantified by the Cox model adjusted for age, sex, smoking, alcohol consumption, exercise, educational level, hypertension, diabetes, hypercholesterolemia, BMI in 2012, hypoglycemic agent use, antihypertensive agent use, lipid-lowering agent use and hs-CRP.
Abbreviations: CVD, cardiovascular disease, AMI, acute myocardial infarction, IS, ischemic stroke, HS, hemorrhagic stroke, HR, hazard ratio.
Among the CVD events, 484 were AMI, 2420 were stroke (2164 were IS and 256 were HS) and 869 were HF. Associations were also observed between WBC count trajectories and increased risks of incident AMI, IS and stroke, with HRs of 1.91 (95% CI: 1.46-2.51), 1.31 (95% CI: 1.16-1.49) and 1.30 (95% CI: 1.16-1.47), respectively, in multivariable-adjusted model 1, the association was nonsignificant for HS (HR=1.23,95% CI: 0.87–1.73) (Supplemental Table 2). Further adjustment for WBC counts in the 2006 and 2012 survey did not markedly change the results (Supplemental Table 2).
| WBC trajectory patterns | |||||
|---|---|---|---|---|---|
| Low- stable | Moderate- stable | Elevated- stable | Moderate- increasing | Elevated- decreasing | |
| AMI | |||||
| Cases, n (incidence rate, per 105 person-years) | 90 (8.23) | 214 (11.87) | 36 (16.83) | 133 (16.91) | 11 (9.00) |
| Multivariate Model 1a | 1 [Reference] | 1.45 (1.13-1.85) | 1.80 (1.21-2.67) | 1.91 (1.46-2.51) | 1.11 (0.59-2.07) |
| Multivariate Model 2b | 1 [Reference] | 1.37 (1.04-1.81) | 1.51 (0.84-2.69) | 1.73 (1.21-2.48) | 0.91 (0.42-1.98) |
| Multivariate Model 3c | 1 [Reference] | 1.24 (0.92-1.67) | 0.98 (0.53-1.79) | 1.44 (0.97-2.16) | 0.85 (0.42-1.74) |
| Stroke | |||||
| Cases, n (incidence rate, per 105 person-years) | 596 (48.44) | 1030 (57.92) | 149 (70.92) | 570 (73.68) | 75 (62.26) |
| Multivariate Model 1a | 1 [Reference] | 1.10 (1.00-1.22) | 1.24 (1.03-1.48) | 1.30 (1.16-1.47) | 1.25 (0.98-1.59) |
| Multivariate Model 2b | 1 [Reference] | 1.08 (0.96-1.22) | 1.16 (0.89-1.50) | 1.26 (1.07-1.47) | 1.16 (0.85-1.60) |
| Multivariate Model 3c | 1 [Reference] | 1.06 (0.94-1.21) | 1.05 (0.80-1.39) | 1.21 (1.02-1.45) | 1.08 (0.82-1.43) |
| IS | |||||
| Cases, n (incidence rate, per 105 person-years) | 524 (42.51) | 930 (52.18) | 137 (65.07) | 508 (65.48) | 65 (53.79) |
| Multivariate Model 1a | 1 [Reference] | 1.13 (1.01-1.26) | 1.28 (1.06-1.55) | 1.31 (1.16-1.49) | 1.23 (0.95-1.59) |
| Multivariate Model 2b | 1 [Reference] | 1.14 (1.01-1.29) | 1.29 (0.98-1.70) | 1.33 (1.12-1.57) | 1.19 (0.85-1.68) |
| Multivariate Model 3c | 1 [Reference] | 1.07 (0.94-1.23) | 1.03 (0.77-1.37) | 1.19 (0.99-1.44) | 1.10 (0.82-1.48) |
| HS | |||||
| Cases, n (incidence rate, per 105 person-years) | 72 (5.85) | 100 (5.62) | 12 (5.71) | 62 (8.01) | 10 (8.3) |
| Multivariate Model 1a | 1 [Reference] | 0.90 (0.66-1.22) | 0.87 (0.47-1.62) | 1.23 (0.87-1.73) | 1.42 (0.73-2.76) |
| Multivariate Model 2b | 1 [Reference] | 0.85 (0.60-1.21) | 0.72 (0.31-1.7) | 1.11 (0.69-1.79) | 1.09 (0.43-2.73) |
| Multivariate Model 3c | 1 [Reference] | 0.98 (0.67-1.43) | 1.31 (0.55-3.09) | 1.43 (0.83-2.48) | 0.90 (0.35-2.31) |
Abbreviations: CVD, cardiovascular disease; AMI, acute myocardial infarction; IS, ischemic stroke; HS, hemorrhagic stroke; HR, hazard ratio.
a Model 1 adjusted for age, sex (male or female), smoking (ever or never), alcohol consumption (ever or never), exercise (＜3 or ≥ 3 times/week), educational level (senior high school or higher or less than senior high school), hypertension (yes or no), diabetes (yes or no), hypercholesterolemia (yes or no), BMI in 2012 (＜24 or ≥ 24 kg/m2), hypoglycemic agent use (yes or no), antihypertensive agent use (yes or no), lipid-lowering agent use (yes or no) and mean serum concentration of high-sensitivity C-reactive protein during 2006-2010 (continuous).
b Model 2 adjusted for the factors in multivariate model 1 plus WBC at 2006, c Model 3 adjusted for the factors in multivariate model 1 plus WBC at 2012.
As of December 31, 2019, a total of 3304 deaths had occurred. The incidence rates in the low-stable pattern, moderate-stable pattern, elevated-stable pattern, moderate-increasing pattern, and elevated-decreasing pattern groups were 66.83, 81.31, 96.03, 104.05 and 66.23 per 105 PYs, respectively. Relative to those in the low-stable pattern group, participants in the elevated-stable pattern group had a significantly higher cumulative rate of all-cause mortality (Fig.3, log-rank test, p＜0.001).
Compared with the low-stable pattern, the elevated-stable pattern appeared to correlate with a higher subsequent risk of all-cause mortality, with HRs of 1.77 (95% CI: 1.52-2.06), 1.56 (95% CI: 1.25-1.96) and 1.35 (95% CI: 1.06-1.73) in multivariable-adjusted models 1-3 (Table 2). In the Fine-Gray model and sensitivity analyses, the moderate-increasing pattern also had a significantly increased risk of all-cause mortality. RCS plots demonstrated significant nonlinear associations of ΔWBC count with all-cause mortality (Fig.4).
WBC Trajectory with Risk of CVD and All-Cause Mortality by CRP ConcentrationNo significant interactions between hs-CRP and WBC trajectories for the risk of CVD and ACM (p-interaction ＞0.05) were observed. The associations of WBC trajectory patterns with CVD and all-cause mortality by hs-CRP concentration (＜2 mg/L, or ≥2 mg/L) are shown in Table 3. In the hs-CRP level ＜2 mg/L or ≥2 mg/L group, compared with the low-stable pattern, the moderate-increasing patterns appeared to similarly correlated with risk of CVD (HR=1.41, 95% CI: 1.24–1.61, HR=1.28, 95% CI: 1.12-1.46, respectively) and ACM (HR=1.44, 95% CI: 1.26–1.66, HR=1.35, 95% CI: 1.16-1.57, respectively).
| WBC trajectory patterns | ||||||
|---|---|---|---|---|---|---|
| Low- stable | Moderate- stable | Elevated- stable | Moderate- increasing | Elevated- decreasing | ||
| P- interaction | ||||||
| CVD | 0.168 | |||||
| Mean hs-CRP≥ 2 mg/L | ||||||
| Cases, n (incidence rate, per 105 person-years) | 406 (97.45) | 791 (110.97) | 138 (133.45) | 464 (134.17) | 60 (107.01) | |
| Multivariate Modela | 1 [Reference] | 1.10 (0.97-1.24) | 1.25 (1.03-1.52) | 1.28 (1.12-1.46) | 1.14 (0.87-1.50) | |
| Mean hs-CRP ＜ 2 mg/L | ||||||
| Cases, n (incidence rate, per 105 person-years) | 505 (63.84) | 846 (80.94) | 93 (89.51) | 426 (102.12) | 44 (69.45) | |
| Multivariate Modela | 1 [Reference] | 1.17 (1.05-1.31) | 1.30 (1.04-1.63) | 1.41 (1.24-1.61) | 1.12 (0.82-1.52) | |
| All-cause mortality | 0.196 | |||||
| Mean hs-CRP≥ 2 mg/L | ||||||
| Cases, n (incidence rate, per 105 person-years) | 408 (97.42) | 773 (107.45) | 154 (147.50) | 451 (129.28) | 50 (87.93) | |
| Multivariate Modela | 1 [Reference] | 1.17 (1.04-1.32) | 1.89 (1.57-2.29) | 1.44 (1.26-1.65) | 1.22 (0.91-1.64) | |
| Mean hs-CRP ＜ 2 mg/L | ||||||
| Cases, n (incidence rate, per 105 person-years) | 413 (51.01) | 670 (63.49) | 64 (60.89) | 291 (68.66) | 30 (46.92) | |
| Multivariate Modela | 1 [Reference] | 1.20 (1.06-1.36) | 1.54 (1.18-2.01) | 1.35 (1.16-1.57) | 1.26 (0.87-1.82) | |
Abbreviations: CVD, Cardiovascular disease, AMI, acute myocardial infarction, IS, ischemic stroke, HS, hemorrhagic stroke, HR, hazard ratio.
a Multivariate Model adjusted for age, sex (men, women), smoking (ever or never smoker), alcohol consumption (ever or never drinker), exercise (＜3 or ≥ 3 times/week), educational level (senior high school and above or less than senior high school), hypertension (yes or no), diabetes (yes or no), hypercholesterolemia (yes or no), BMI in 2012 (＜24 or ≥ 24 kg/m2), hypoglycemic agent use (yes or no), antihypertensive agent use (yes or no), lipid-lowering agent use (yes or no).
Although he highest risk of AMI was observed in those with the moderate-increasing pattern among the subgroup with hs-CRP ＜2 mg/L (HR=2.44, 95% CI: 1.63-3.64), the interaction between hs-CRP and WBC trajectories for the AMI were not also observed (P=0.056, for interaction). (Supplemental Table 3).
| the WBC trajectory patterns | |||||
|---|---|---|---|---|---|
| Low- stable | Moderate- stable | Elevated- stable | Moderate- increasing | Elevated- decreasing | |
| AMI | |||||
| Mean hs-CRP ≥ 2 mg/L | |||||
| Cases, n (incidence rate, per 105 person-years) | 50 (11.72) | 116 (15.78) | 22 (20.48) | 69 (19.27) | 5 (8.65) |
| Multivariate Modela | 1 [Reference] | 1.26 (0.9-1.75) | 1.51 (0.91-2.5) | 1.44 (1.00-2.09) | 0.70 (0.28-1.75) |
| Mean hs-CRP＜2 mg/L | |||||
| Cases, n (incidence rate, per 105 person-years) | 40 (4.89) | 98 (9.18) | 14 (13.14) | 64 (14.94) | 6 (9.31) |
| Multivariate Modela | 1 [Reference] | 1.72 (0.73-4.08) | 2.16 (1.17-4.00) | 2.44 (1.63-3.64) | 1.64 (1.13-2.37) |
AMI, acute myocardial infarction; HR, hazard ratio.
a Multivariate Model adjusted for age, sex (male or female), smoking (ever or never), alcohol consumption (ever or never), exercise (＜3 or ≥ 3 times/week), educational level (senior high school or higher or less than senior high school), hypertension (yes or no), diabetes (yes or no), hypercholesterolemia (yes or no), BMI in 2012 (＜24 or ≥ 24 kg/m2), hypoglycemic agent use (yes or no), antihypertensive agent use (yes or no), lipid-lowering agent use (yes or no).
In this prospective cohort study, five heterogeneous WBC trajectories were identified in a homogenous Chinese population, and these patterns were related to incident CVD and all-cause mortality. Compared with those with a low-stable WBC pattern, individuals with a moderate-increasing WBC pattern had a higher risk for incident CVD, even after adjustment for single WBC count measurements. In addition, elevated-stable WBC patterns were linked to an increased risk of all-cause mortality. The study initially discovered a positive association between WBC trajectories and the risks for CVD and all-cause mortality.
Based on repeated WBC measurements from the Kailuan cohort, we discovered distinct WBC trajectories that are similar to those of blood pressure25) and glucose30). In this cohort, 73.12% of the participants had normal WBC counts, with a slight decrease with age. This was consistent with the Baltimore longitudinal aging study17). However, 23.96% of individuals maintained high WBC counts (elevated-stable pattern) or appeared to be on an increasing trajectory (moderate-increasing pattern). Furthermore, individuals with an increasing trajectory were more likely to have a concurrent CVD risk factor than those in the low-stable pattern group.
The present analyses showed that participants with moderate-increasing WBC patterns had a higher CVD risk. After adjustment for age, sex, hypertension, diabetes, hypercholesterolemia, hs-CRP, and other cardiovascular risk factors, the moderate-increasing pattern was associated with a 36% higher CVD risk than the low-stable WBC pattern. This association demonstrated that the CVD risk due to an increasing WBC count is independent of traditional risk factors and medication. On the other hand, we did not find a positive association between the elevated-decreasing pattern and CVD risk, although WBC counts in the elevated-decreasing pattern group were higher than those in the moderate-increasing pattern group (9.9×109/L vs. 7.6×109/L) in the first measurement. These findings suggested that an increased risk of CVD associated with a higher WBC count could be attenuated as long as the WBC count decrease over time and demonstrated the advantage of the trajectory model over single measurement. To our knowledge, no similar studies have been conducted. However, a possible link between dynamic WBC changes and CVD risk has been suggested in a few previous studies based on two WBC measurements. According to the Dongfeng-Tongji study31), increases in WBC counts at 5-year intervals in elderly individuals increased the risks of CVD and stroke, whereas decreases in WBC counts in the MERIFT study32) and the MELANY study33) reduced the risks of AMI and coronary calcification, respectively. In contrast to previous research, we directly demonstrated a positive correlation between WBC variations and CVD risk in the general population.
Another interesting finding of our study is that an association between WBC trajectories and the risks of incident CVD, MI and all-cause mortality was not accordance with serum hs-CRP level. The findings of stratified analyses revealed that there was the associations of WBC trajectories with all-cause mortality were similar between the groups above versus below CRP ＞2 mg/L This implies that in conditions of low inflammation(hs-CRP ＜2 mg/L), increasing WBC count still raise the risk of CVD and ACM. We speculate a major factor is aberrant immunological activation. WBC count is an index of immune status34). An increase in the WBC count might suggest immune activation in the presence of CVD35). Animal experiments have shown that auto-antibodies, such as anti-oxLDL and anti-p100, and auto-reactive lymphocytes, such as CD8＋ T cells and B2 cells, can induce macrophage and vascular smooth muscle cell apoptosis in plaques, accelerate the formation of necrotic cores and promote plaque instability36-38). Clinical data also support that auto-antibodies and auto-reactive lymphocytes increase the risk of coronary stenosis and AMI39, 40). In addition, two other factors could be involved. First, abnormal conditions of the Hematopoietic system, may also lead to changed of counts of blood cells and an increased risk of CVD. For example, clonal hematopoiesis of indeterminate potential was found to increase the risk of CVD41). Second, the silent exposure, such as viral infection, tuberculosis, periodontal disease, heavy metal exposure, air pollution, etc., may not cause CRP elevations that lead to mild leukocyte elevations, and also related with the risk of CVD events.
The results also showed an increased risk of all-cause mortality in the moderate-increasing stable pattern group but not in the elevated-decreasing pattern group. This finding demonstrates the adverse impacts of chronic inflammation and aberrant immune activation on health, indicated by elevated WBC counts, and these impacts might be reduced by lowering the elevated WBC. Interestingly, the effects of the WBC trajectory on all-cause mortality and CVD risks were not simultaneous. The risk for all-cause mortality, but not for CVD, increased in the elevated-stable pattern group. This could be attributed to the following factors. First, the risk of CVD in the elevated-stable pattern group may be underestimated due to the competing risk of death. Second, the proportion of those who used lipid-lowering medicines, hypoglycemic drugs, or antihypertensive drugs was higher in the high stable group. These medications can lower the risk of CVD but do not reduce the risk of death from other causes. Finally, patients with a high WBC count are more likely to develop cancer13, 17). The group with the high stable pattern also had more malignant tumor risk factors, such as smoking, alcohol consumption, and a higher BMI.
WBC counts affect CVD and death via complex mechanisms, the most important of which is the atherogenic effect. Leukocytes can initiate the atherosclerotic process by differentiating vascular smooth muscle cells and activating immune responses and other actions42). In addition, WBCs are involved in the rupture and thrombosis of unstable atherosclerotic plaques43). Increased WBC counts and conversion to a proinflammatory phenotype induced by hypercholesterolemia44, 45) can promote atherosclerotic enlargement and vulnerable plaque formation and accelerate the CVD process. In addition, the chronic inflammatory response represented by elevated WBC counts is a marker of aging, and degenerative pathologies such as Alzheimer’s disease, tumors, and osteoarthritis are important causes of death.
The strengths of this study are the large sample size, the prospective design, and the application of trajectory models. We found a positive association between dynamic changes in WBC count and the risk of CVD and death, demonstrating the advantages of WBC trajectories in the evaluation of CVD and all-cause mortality risks.
The present study also has some limitations. First, many factors affect WBC counts. Although confounding factors could not be completely excluded, we accounted for diseases that have an effect on WBC counts and outliers that may be due to infection, moreover, multiple repeated measurements over 8 years ensured the reliability of the WBC trajectory. Second, the population in this study was predominantly northern Chinese males, and the results may not apply to populations from other regions and ethnicities. Finally, this study was observational, and the findings do not thoroughly confirm the causal relationship between dynamic WBC trajectories and endpoint events.
In conclusion, five distinct WBC count trajectories were identified in our study. There might be direct relationships between an increasing WBC count trajectory and CVD and all-cause mortality. Monitoring WBC count trajectories may be an important approach to identify populations at increased risk of CVD and all-cause mortality and may help to prevent primary CVD in the general population. Studies in populations with different racial and ethnic compositions and those that include data on different components of WBC counts are also warranted to validate our findings.
The authors thank the participants in the Kailuan Study, the laboratory staff of the Department of Cardiology of Kailuan Hospital, and the staff of the Department of Hematology of the General Hospital of Tianjin Medical University.
Not applicable.
All the authors declare that they have no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Trial identification: ChiCTR-TNC-11001489,
Registration site: http://www.chictr.org.cn/index.aspx, Registration number: 11001489
